Hi, I am using linaro 13.03 image with ARMv8 fast model. I am getting following "unhandled L1 translation fault" exception while executing a test. Is it a known problem? Does anybody know its solution? I am getting the same error while executing perl benchmark from SPEC 2000.
Thanks, Aparna
[[ 364.733273] jsc[834]: unhandled level 1 translation fault (11) at 0xffffffff874c0120, esr 0x92000005 [ 364.733356] pgd = ffffffc87d22b000 [ 364.733421] [ffffffff874c0120] *pgd=00000008ffeb9003, *pmd=00000008fe78f003, *pte=00a00008fbd74f53 [ 364.733573] [ 364.733631] Pid: 834, comm: jsc [ 364.733716] CPU: 0 Not tainted (3.9.0-1-linaro-vexpress64 #1ubuntu1~ci+130320041336) [ 364.733796] PC is at 0x4f69d4 [ 364.733860] LR is at 0x44cdec [ 364.733933] pc : [<00000000004f69d4>] lr : [<000000000044cdec>] pstate: 60000000 [ 364.734008] sp : 0000007ff1c891f0 [ 364.734070] x29: 0000007ff1c891f0 x28: 0000000000000000 [ 364.734188] x27: 0000007ff1c89db8 x26: 00000000005258a0 [ 364.734308] x25: 0000007f89342fa8 x24: 0000007f874c00c0 [ 364.734428] x23: 0000007ff1c89900 x22: 000000003f66ea48 [ 364.734548] x21: 0000007f89342e00 x20: 0000007f8933a4b0 [ 364.734669] x19: ffffffff874c0100 x18: 0000007ff1c899a0 [ 364.734789] x17: 0000007f8a2ac7d4 x16: 00000000005827b8 [ 364.734908] x15: 0000007f8a0ca590 x14: 0000000000000000 [ 364.735026] x13: 000000000000001f x12: 0000000000000000 [ 364.735144] x11: 0000000000000000 x10: 0000007f89345420 [ 364.735262] x9 : 000000000000000f x8 : 000000000000000f [ 364.735378] x7 : 000000000000000f x6 : 0000007f89353ca0 [ 364.735498] x5 : 0000007f893433e8 x4 : 0000007f89353cc0 [ 364.735618] x3 : 0000007f89353c40 x2 : 0000000000000003 [ 364.735735] x1 : 000000003f66ea48 x0 : 0000000000000000 [ 364.735845]
On 9 May 2013 09:26, Aparna Mandke aparna.mandke@gmail.com wrote:
I am using linaro 13.03 image with ARMv8 fast model. I am getting following "unhandled L1 translation fault" exception while executing a test. Is it a known problem? Does anybody know its solution? I am getting the same error while executing perl benchmark from SPEC 2000.
...
[[ 364.733273] jsc[834]: unhandled level 1 translation fault (11) at 0xffffffff874c0120, esr 0x92000005
...
[ 364.733796] PC is at 0x4f69d4 [ 364.733860] LR is at 0x44cdec
It looks like an application (jsc) is trying to access some kernel memory directly. Why it does this, I don't know. You can disassemble instructions around the PC value above.
Catalin
Following is the disassembly obtained using objdump: 4f69d4: b9402260 ldr w0, [x19,#32]
Thanks, Aparna
On Thu, May 9, 2013 at 2:09 PM, Catalin Marinas catalin.marinas@arm.comwrote:
On 9 May 2013 09:26, Aparna Mandke aparna.mandke@gmail.com wrote:
I am using linaro 13.03 image with ARMv8 fast model. I am getting following "unhandled L1 translation fault" exception while executing a test. Is it a known problem? Does anybody know its solution?
I
am getting the same error while executing perl benchmark from SPEC 2000.
...
[[ 364.733273] jsc[834]: unhandled level 1 translation fault (11) at
0xffffffff874c0120, esr 0x92000005 ...
[ 364.733796] PC is at 0x4f69d4 [ 364.733860] LR is at 0x44cdec
It looks like an application (jsc) is trying to access some kernel memory directly. Why it does this, I don't know. You can disassemble instructions around the PC value above.
Catalin
Printing disassembly of more instructions in this region:
4f6988: a9bc7bfd stp x29, x30, [sp,#-64]! 4f698c: 910003fd mov x29, sp 4f6990: a90153f3 stp x19, x20, [sp,#16] 4f6994: f90013f5 str x21, [sp,#32] 4f6998: aa0003f3 mov x19, x0 4f699c: f9400674 ldr x20, [x19,#8] 4f69a0: f85c8020 ldr x0, [x1,#-56] 4f69a4: f9402283 ldr x3, [x20,#64] 4f69a8: f9400800 ldr x0, [x0,#16] 4f69ac: f9405415 ldr x21, [x0,#168] 4f69b0: b4000703 cbz x3, 4f6a90 <_ZN3JSC16InternalFunction4nameEPNS_9ExecStateE+0x108> 4f69b4: f9409ea0 ldr x0, [x21,#312] 4f69b8: b9400067 ldr w7, [x3] 4f69bc: b9401802 ldr w2, [x0,#24] 4f69c0: 0a070044 and w4, w2, w7 4f69c4: 8b244864 add x4, x3, w4, uxtw #2 4f69c8: b9402084 ldr w4, [x4,#32] 4f69cc: 35000164 cbnz w4, 4f69f8 <_ZN3JSC16InternalFunction4nameEPNS_9ExecStateE+0x70> 4f69d0: d2800013 mov x19, #0x0 // #0 4f69d4: b9402260 ldr w0, [x19,#32] 4f69d8: 34000060 cbz w0, 4f69e4 <_ZN3JSC16InternalFunction4nameEPNS_9ExecStateE+0x5c> 4f69dc: aa1303e0 mov x0, x19 4f69e0: 97fd8b72 bl 4597a8 <_ZNK3JSC8JSString11resolveRopeEPNS_9ExecStateE> 4f69e4: 91006260 add x0, x19, #0x18 4f69e8: f94013f5 ldr x21, [sp,#32] 4f69ec: a94153f3 ldp x19, x20, [sp,#16] 4f69f0: a8c47bfd ldp x29, x30, [sp],#64
On Thu, May 9, 2013 at 2:50 PM, Aparna Mandke aparna.mandke@gmail.comwrote:
Following is the disassembly obtained using objdump: 4f69d4: b9402260 ldr w0, [x19,#32]
Thanks, Aparna
On Thu, May 9, 2013 at 2:09 PM, Catalin Marinas catalin.marinas@arm.comwrote:
On 9 May 2013 09:26, Aparna Mandke aparna.mandke@gmail.com wrote:
I am using linaro 13.03 image with ARMv8 fast model. I am getting following "unhandled L1 translation fault" exception while executing a test. Is it a known problem? Does anybody know its
solution? I
am getting the same error while executing perl benchmark from SPEC 2000.
...
[[ 364.733273] jsc[834]: unhandled level 1 translation fault (11) at
0xffffffff874c0120, esr 0x92000005 ...
[ 364.733796] PC is at 0x4f69d4 [ 364.733860] LR is at 0x44cdec
It looks like an application (jsc) is trying to access some kernel memory directly. Why it does this, I don't know. You can disassemble instructions around the PC value above.
Catalin
looks like some other issue. 4f69d0: d2800013 mov x19, #0x0 // #0 4f69d4: b9402260 ldr w0, [x19,#32]
So it is trying to access - 0x32 address and that is why L1 translation fault. Problem is somewhere else. Thanks Aparna
On Thu, May 9, 2013 at 2:53 PM, Aparna Mandke aparna.mandke@gmail.comwrote:
Printing disassembly of more instructions in this region:
4f6988: a9bc7bfd stp x29, x30, [sp,#-64]! 4f698c: 910003fd mov x29, sp 4f6990: a90153f3 stp x19, x20, [sp,#16] 4f6994: f90013f5 str x21, [sp,#32] 4f6998: aa0003f3 mov x19, x0 4f699c: f9400674 ldr x20, [x19,#8] 4f69a0: f85c8020 ldr x0, [x1,#-56] 4f69a4: f9402283 ldr x3, [x20,#64] 4f69a8: f9400800 ldr x0, [x0,#16] 4f69ac: f9405415 ldr x21, [x0,#168] 4f69b0: b4000703 cbz x3, 4f6a90 <_ZN3JSC16InternalFunction4nameEPNS_9ExecStateE+0x108> 4f69b4: f9409ea0 ldr x0, [x21,#312] 4f69b8: b9400067 ldr w7, [x3] 4f69bc: b9401802 ldr w2, [x0,#24] 4f69c0: 0a070044 and w4, w2, w7 4f69c4: 8b244864 add x4, x3, w4, uxtw #2 4f69c8: b9402084 ldr w4, [x4,#32] 4f69cc: 35000164 cbnz w4, 4f69f8 <_ZN3JSC16InternalFunction4nameEPNS_9ExecStateE+0x70> 4f69d0: d2800013 mov x19, #0x0 // #0 4f69d4: b9402260 ldr w0, [x19,#32] 4f69d8: 34000060 cbz w0, 4f69e4 <_ZN3JSC16InternalFunction4nameEPNS_9ExecStateE+0x5c> 4f69dc: aa1303e0 mov x0, x19 4f69e0: 97fd8b72 bl 4597a8 <_ZNK3JSC8JSString11resolveRopeEPNS_9ExecStateE> 4f69e4: 91006260 add x0, x19, #0x18 4f69e8: f94013f5 ldr x21, [sp,#32] 4f69ec: a94153f3 ldp x19, x20, [sp,#16] 4f69f0: a8c47bfd ldp x29, x30, [sp],#64
On Thu, May 9, 2013 at 2:50 PM, Aparna Mandke aparna.mandke@gmail.comwrote:
Following is the disassembly obtained using objdump: 4f69d4: b9402260 ldr w0, [x19,#32]
Thanks, Aparna
On Thu, May 9, 2013 at 2:09 PM, Catalin Marinas catalin.marinas@arm.comwrote:
On 9 May 2013 09:26, Aparna Mandke aparna.mandke@gmail.com wrote:
I am using linaro 13.03 image with ARMv8 fast model. I am getting following "unhandled L1 translation fault" exception while executing a test. Is it a known problem? Does anybody know its
solution? I
am getting the same error while executing perl benchmark from SPEC
...
[[ 364.733273] jsc[834]: unhandled level 1 translation fault (11) at
0xffffffff874c0120, esr 0x92000005 ...
[ 364.733796] PC is at 0x4f69d4 [ 364.733860] LR is at 0x44cdec
It looks like an application (jsc) is trying to access some kernel memory directly. Why it does this, I don't know. You can disassemble instructions around the PC value above.
Catalin
On Thu, May 09, 2013 at 10:27:53AM +0100, Aparna Mandke wrote:
looks like some other issue. 4f69d0: d2800013 mov x19, #0x0 // #0 4f69d4: b9402260 ldr w0, [x19,#32]
So it is trying to access - 0x32 address and that is why L1 translation fault. Problem is somewhere else.
Can you reproduce this easily? Is it an application that you run and crashes consistently?
The strange thing is that the fault address should be reported as 32 rather than some high 0xffffffff874c0120 value. In your kernel message, x19 is indeed 0xffffffff874c0100 which does not match the previous mov x19, #0.
Hi Catalin,
disassembly at 0x4f69d0 does not look right though opcode is same. 4f69d0: d2800013 movz x19, #0x0 4f69d4: b9402260 ldr w0, [x19,#32]
It is movz and not just mov. So I think data contents of x19 are okay. I can reproduce this bug consistently. Thanks, Aparna
44cdc4: a9bc7bfd stp x29, x30, [sp,#-64]!
44cdc8: 910003fd mov x29, sp
44cdcc: a90153f3 stp x19, x20, [sp,#16]
44cdd0: aa0003f3 mov x19, x0
44cdd4: aa0203e0 mov x0, x2
44cdd8: a9025bf5 stp x21, x22, [sp,#32]
44cddc: aa0203f4 mov x20, x2
44cde0: aa0103f6 mov x22, x1
44cde4: 2a0303f5 mov w21, w3
44cde8: 9402a6e8 bl 4f6988 JSC::InternalFunction::name(JSC::ExecState*) 44cdec: f9400001 ldr x1, [x0] 44cdf0: b9400022 ldr w2, [x1]
44cdf4: 36100522 tbz w2, #2, 44ce98 <JSC::JSObject::putDirectFunctionWithoutTransition(JSC::ExecState*, JSC::InternalFunction*, unsigned int)+0xd4> 44cdf8: f9001ba1 str x1, [x29,#48]
44cdfc: b9400020 ldr w0, [x1]
44ce00: 11008000 add w0, w0, #0x20
On Thu, May 9, 2013 at 4:01 PM, Catalin Marinas catalin.marinas@arm.comwrote:
On Thu, May 09, 2013 at 10:23:23AM +0100, Aparna Mandke wrote:
Printing disassembly of more instructions in this region:
Can you also get a disassemble around LR (0x44cdec)?
-- Catalin
On Thu, May 09, 2013 at 05:52:32PM +0100, Aparna Mandke wrote:
Hi Catalin,
disassembly at 0x4f69d0 does not look right though opcode is same. 4f69d0: d2800013 movz x19, #0x0 4f69d4: b9402260 ldr w0, [x19,#32]
It is movz and not just mov. So I think data contents of x19 are okay.
mov #imm is an alias for movz. So after movz, x19 should be 0, in which case the subsequent ldr does not make sense.
I can reproduce this bug consistently.
Is the kernel printing the same value for x19 consistently?
linaro-kernel@lists.linaro.org
-
Aparna Mandke -
Catalin Marinas